primeline-ai/evolving-lite/system-boot

System Boot

On every session start, perform this bootup sequence:

1. Read Memory

Read ${CLAUDE_PLUGIN_ROOT}/_memory/index.json for:

• active_project: Which project the user was working on
• last_session: When the last session was

If active_project is set, read ${CLAUDE_PLUGIN_ROOT}/_memory/projects/{active_project}.json for:

• progress: Array of recent progress entries
• next_step: What was planned next
• failures: Known blockers or failures

2. Orient

Based on memory:

• What was the last progress entry?
• Are there any known failures or blockers?
• What was suggested as the next step?

3. Announce

Output a compact status line:

Evolving Lite | Session {n} | Tier {tier} | {experience_count} experiences
Last: {last_progress_summary}
Next: {next_step}

If this is the first session ever (no progress), announce with a short, visible cold-start sequence so the system does not feel inert on day 1:

Evolving Lite v1.0 | First session
  [1/3] Loading {prewarmed_count} pre-warmed experiences ... ok
  [2/3] Tier 1 (Safety) active: context warnings, bash security, hook sentinels
  [3/3] Self-Star Doctor: wiring verified (run /health for the full board)
Ready. The system learns from your corrections automatically and self-tunes its
delegation routing from your sessions (see "Self-Evolution is ON" in the README).

The Self-Star Doctor also runs a quick wiring + preflight check automatically on this first session. If it reports a missing prerequisite (most often Kairn), surface that to the user and point them at /health and pip install kairn-ai.

4. Pick One Task

If the user hasn't specified what to work on, suggest the next_step from memory. Don't start multiple tasks - pick one.

Continue Trigger

When the user says "continue", "weiter", "weitermachen", "fortsetzen", or "resume":

1. Read the most recent session summary from ${CLAUDE_PLUGIN_ROOT}/_memory/sessions/
2. Load the plan if one was referenced
3. Continue immediately - no questions, just pick up where we left off

Session End

When a session ends naturally (user says goodbye, or Stop hook fires):

• The session-summary hook handles progress logging automatically
• No manual action needed from the user